Functional brain connectivity of audiovisual speech processing using graph filtration

Abstract

Several brain regions have been implicated in audiovisual speech integration. However, the functional network that exists between these regions to facilitate multisensory speech integration remains unclear. Previous studies suggest that the superior temporal sulcus/gyrus (STS/STG) is a critical brain area for multisensory integration, but functional connectivity between STS/STG and other regions involved in audiovisual speech processing has not been observed consistently, perhaps in part due to differences in the kinds of tasks or thresholds used across studies. To avoid this issue, the current study used graph filtration in a persistent homology framework to investigate the functional networks among brain areas activated during audiovisual speech processing. The audiovisual speech task in this study included 4 conditions that differed in the sensory modality used to deliver speech: audiovisual speech (AV), auditory speech (A), visual speech (V), and audiovisual non-speech (C). I constructed three speech-specific brain networks that were represented by barcode and minimum spanning tree using single linkage distance and compared functional networks. These results revealed that the audiovisual speech network was focused on connectivity between the posterior STG (pSTG) and inferior frontal region, while the visual speech network was focused on the connectivity between the hippocampal and inferior frontal regions. Prominent functional connectivity was observed between the pSTG and frontal regions during audiovisual speech integration, and functional connectivity between the temporal regions and auditory or visual areas was driven by auditory or visual speech modalities. Additionally, meaningful visual speech without auditory speech seems to depend on the retrieval of previous memories related to speech production. The results of this study suggest that the mechanisms by which the brain uses visual information to improve auditory speech perception may involve fronto-temporal connectivity with motor areas related to speech production, rather than visual sensory areas.